and the and the Ponidicherry Conference, 2010 The University of Iowa goodman@math.uiowa.edu
and the The goal of this work is to certain finite dimensional algebras that arise in invariant theory, knot theory, subfactors, QFT, and statistical mechanics. The algebras in question have parameters; for generic values of the parameters, they are semisimple, but it is also interesting to non semisimple specializations. It turns out that certain ideas from the semisimple world specifically, the are still useful in the non semisimple case. The
The There are two themes in this work: " and the." means something specific, and it will be defined. What the means in the general context we consider is not exactly clear. When it does make sense, the is a machine which, given a pair of algebras 1 A B, will produce a third algebra J, with A B J. If A and B are split semisimple over a field F, then it is clear what J should be, namely J = End(B A ). Suppose now that A and B are not only split s.s., but also that we have an F valued trace on B, which is faithful on B and has faithful restriction to A. (Faithful means that the bilinear form (x, y) = tr(xy) is non degenerate.) and the The
Basic, cont. and the In the case just described, we have a unital A A bimodule map ɛ : B A determined by tr(ba) = tr(ɛ(b)a) for b B and a A. Then we also have End(B A ) = BɛB = B A B, where the latter isomorphism is as B B bimodules Now consider three successive A n 1 A n A n+1. Then A n+1 contains an essential idempotent e n, and we would like to understand the ideal J n+1 = A n+1 e n A n+1. The
Basic, slide 3 Suppose (for the moment) that we work over and the parameters of the are chosen generically. Then all the algebras in sight are s.s. and J n+1 = End((An ) An 1 ) = A n An 1 A n. That is, J n+1 is the basic for A n 1 A n. Does some part of this persist in the non s.s. case? Now, work over the generic integral ground ring for the BMW algebras; i.e. work with the integral form" of the BMW algebras. Now J n+1 is no longer even a unital algebra and A n is not projective as an A n 1 module. Nevertheless, it remains true that J n+1 = An An 1 A n. and the The This is what makes our treatment of cellularity for BMW algebras, and other similar algebras, work.
and the The is a concept due to Graham and Lehrer that is useful in the of a number of important algebras: Hecke algebras, Brauer and, Schur algebras, etc. I will give the, say what it is good for, and then make some general observations about cellularity, including a suggested variant on the, and a new basis free formulation.
What is cellularity? Let A be an algebra with involution over an integral domain S. A is said to be cellular if there exists a finite partially ordered set (Λ, ) and for each λ Λ, a finite index set (λ), such that A has an S basis {cs,t λ : λ Λ;s,t (λ)}. (c λ s,t ) = c λ t,s. For a A, a c λ s,t s r c λ r,t, modulo the span of basis elements c µ u,v, with µ > λ, where the coefficients in the expansion depend only on a and s and not on t. r and the The Such a basis is called a cellular basis, and the whole apparatus (Λ,, (λ),{cs,t λ }) is called a cell datum.
Example Definition 1 The Hecke algebra H S n (q2 ) over a ring S is the quotient of the braid group algebra over S by the Hecke skein relation: = (q q 1 ). Fact: The Hecke algebras H S n (q2 ) are cellular, with Λ = Y n, the set of Young diagrams with n boxes, ordered by dominance, and (λ) the set of standard tableaux of shape λ. The cell modules are known as Specht modules. The cellular structure is due to Murphy. and the The See, for example, A. Mathas, Iwahori-Hecke Algebras and Schur Algebras of the Symmetric Group, AMS University Lecture Series.
What is cellularity?, cont. It follows immediately from the of cellularity that For every order ideal Γ of Λ, is a ideal of A. A(Γ) := span{c λ s,t : λ Γ,s,t (λ)} In particular, write A λ for A({µ : µ λ}) and Ă λ for A({µ : µ > λ}). For each λ Λ, there is an A module λ, free as S module, with basis {ct λ : t (λ)}, such that the map α λ : A λ /Ă λ λ R ( λ ) defined by α λ : cs,t λ + Ăλ c s (ct λ) is an A A bimodule isomorphism. and the The
What is this good for? and the The modules λ are called cell modules. When the ground ring is a field, and the algebra A is semisimple, these are exactly the simple modules. In general, general, λ has a canonical bilinear form. With rad(λ) the radical of this form, and with the ground ring a field, λ /rad(λ) is either zero or simple, and all simples are of this form. So cellularity gives an approach to finding all the simple modules. It s also useful for investigating the block structure, etc. The
Proposed weakening of the and the I propose weakening the of cellularity, replacing by (c λ s,t ) = c λ t,s (c λ s,t ) c λ t,s modulo Ăλ. Disadvantages: none that I know of. All the results of Graham & Lehrer are still valid. Moreover, the weakened is equivalent to the original if 2 is invertible in the ground ring, so we haven t lost much. The
Advantages of the weaker and the Main advantage is graceful treatment of extensions with weaker : If J is a cellular ideal" in A and A/J is cellular, then A is cellular. Cellular algebras can have many different cellular bases yielding the same cellular structure," i.e. same ideals, same cell modules. Suppose one has a cellular algebra with cell modules λ. With the weaker, one easily sees that any collection of bases of the cell modules can be lifted (globalized) to a cellular basis of the algebra, via the maps α λ : A λ /Ă λ λ R ( λ ) The
With the weakened of cellularity, one gets a basis free formulation of cellularity (improving on work of König & Xi). In the following, a Λ cell net" is a map from order ideals of Λ to ideals of A, with several natural properties. Denote the map Γ J(Γ). The most important properity is: for each λ Λ, there exists an A module M λ, free as S module, such that J(Γ)/J(Γ ) = M λ S (M λ ), as A A bimodules, whenever Γ \ Γ = {λ}. Proposition 2 Let A be an R algebra with involution, and let (Λ, ) be a finite partially ordered set. Then A has a cell datum with partially ordered set Λ if, and only if, A has a Λ cell net. Moreover, in this case, the M λ are the cell modules for the cellular structure. and the The
Our objects of We develop a framework for ing cellularity for several important examples of pairs of towers of algebras, such as A 0 A 1 A 2... and Q 0 Q 1 Q 2... A n = Brauer algebra, Q n = symmetric group algebra. A n = BMW algebra, Q n = Hecke algebra. A n = cyclotomic BMW algebra, Q n = cyclotomic Hecke algebra. A n = partition algebra, Q n = stuttering" sequence of symmetric group algebras. etc. and the The
Some properties of the examples and the 1. There is a generic (integral) ground ring R for A n (independent of n). Every instance of A n over a ground ring S is a specialization: A S n = AR n R S. 2. With F the field of fractions of R, the algebra A F n is semisimple. 3. A n contains an essential idempotent e n 1. Set J n = A n e n 1 A n. Then: Q n = An /J n, and Jn+1 F is isomorphic to a for the pair A F n 1 AF n. All this results from a method due to Wenzl in the 80 s for showing the generic semisimplicity and determining the generic branching diagram of the tower (A F n ) n 0. (Wenzl applied this to Brauer and.) The
What we did and the The In this work we have found a cellular analogue of Wenzl s, which gives a uniform of cellularity of the algebras A n in the examples, with additional desirable features. Note that for results about cellularity, it suffices for us to work over the generic ground ring R, since cellularity is preserved under specialization.
and the The It is a general principle that representation theories of the Hecke algebras H n (q) or of the symmetric group algebras KS n should be considered all together, that induction/restriction between H n and H n 1 plays a role in building up the representation theory. is the cellular version of this principle.
Coherence and strong coherence and the Definition 3 A sequence (A n ) n 0 of cellular algebras, with cell data (Λ n,...) is coherent if for each µ Λ n, the restriction of µ to A n 1 has a filtration Res( λ ) = F t F t 1 F 0 = (0), with F j /F j 1 = λ j for some λ j Λ n 1, and similarly for induced modules. The sequence (A n ) n 0 is strongly coherent if, in addition, λ t < λ t 1 < < λ 1 The in Λ n 1, and similarly for induced modules.
Strong coherence and path bases and the Let (A n ) n 0 be a strong coherent sequence of cellular algebras over R, and suppose the algebras A F n over the field of fractions F of R are (split) semisimple. Then the simple modules for A F n have bases indexed by paths on the branching diagram (Bratteli diagram) B for the sequence (A k ) 0 k n. It is natural to want bases of the cell modules of A n also indexed by paths on B and having good properties with respect to restriction. Call such bases of the cell modules and their globalization to cellular bases of A n path bases With mild additional assumptions, satisfied in our examples, one always has path bases in strongly coherent towers. The
Example of strong coherence and the The The sequence of Hecke algebras H n (q) is a strongly coherent sequence of cellular algebras, and the Murphy basis is a path basis. This results from combining theorems of Murphy, Dipper-James, and Jost from the 80 s
Main theorem Theorem 4 Suppose (A n ) n 0 and (Q n ) n 0 are two sequence of - algebras over R. Let F be the field of fractions of R. Assume: 1. (Q n ) n 0 is a (strongly) coherent tower of cellular algebras. 2. A 0 = Q 0 = R, A 1 = Q1. 3. For each n 2, A n has an essential idempotent e n 1 = e n 1 and A n/a n e n 1 A n = Qn. 4. A F n = A n R F is split semisimple. 5. (a) e n 1 commutes with A n 1, and e n 1 A n 1 e n 1 A n 2 e n 1, (b) A n e n 1 = A n 1 e n 1 and x xe n 1 is injective from A n 1 to A n 1 e n1, and (c) e n 1 = e n 1 e n e n 1. Then (A n ) n 0 is a (strongly) coherent tower of cellular algebras. and the The
Example: The A chief example is A n = BMW (Birman-Murakami-Wenzl) algebra on n strands, and Q n = Hecke algebra on n strands. arise in knot theory (Kauffman link invariant) and in quantum invariant theory (Schur Weyl duality for quantum groups of symplectic and orthogonal types). The BMW algebra is an algebra of braid like objects, namely framed (n, n) tangles :. Tangles can be represented by quasi-planar diagrams as shown here. Tangles are multiplied by stacking (like braids). and the The
Definition of Definition 5 Let S be a commutative unital ring with invertible elements ρ, q,δ satisfying ρ 1 ρ = (q 1 q)(δ 1). The BMW algebra Wn S is the S algebra of framed (n, n) tangles, modulo the Kauffman skein relations: 1. (Crossing relation) = (q 1 q) 2. (Untwisting relation) = ρ and = ρ 1. 3. (Free loop relation) T = δt, where T is the union of a tangle T and an additional closed loop with zero framing.. and the The
The, cont. and the Wn S imbeds in W n+1 S. On the level of tangle diagrams, the embedding is by adding one strand on the right. The have an S linear algebra involution, acting by turning tangle diagrams upside down. The following tangles generate the BMW algebra The e i = i i + 1 and g i = i +1 The element e i is an essential idempotent with e 2 i = δe i. One has e i e i±1 e i = e i. i.
The, cont. 2 and the The ideal J n generated by one or all e i s in W S n satisfies W S n /J n = H S n (q2 ) where H S n (q2 ) is the Hecke algebra. There is a generic ground ring for the, namely R = [q ±1,ρ ±1,δ]/J, where J is the ideal generated by ρ 1 ρ (q 1 q 1 )(δ 1) and where the bold symbols denote indeterminants. The
BMW Generic ground ring, cont. 3 and the The The generic ground ring R is an integral domain, with field of fractions F = (q,ρ), and δ = (ρ 1 ρ)/(q 1 q) + 1 in F. For every instance of the BMW over a ring S with parameters ρ, q,δ, one has W S n = W R n R S. The over F are semisimple (theorem of Wenzl).
The, application of our theorem and the Now let s see what s involved in applying the theorem to the (with A n = W R n, and Q n = H R n (q2 )). Hypothesis (1) is the (strong) coherence of the sequence of Hecke algebras, which is a significant theorem about Hecke algebras. Hypothesis (4) on the semisimplicity of W F n is Wenzl s theorem. Everything else is elementary, and already contained in Birman-Wenzl. All the other examples work pretty much the same way. The
and the The is inductive and is a cellular version of Wenzl s semisimplicity. Suppose we know that A k is cellular (and satisfies all the conclusions of the theorem) for k n. Then we want to show the same for A n+1. The main point is to show that J n+1 = A n+1 e n A n+1 = A n e n A n is a cellular ideal" in A n+1. (This suffices to show cellularity, because we also have that A n+1 /J n+1 = Qn+1 is cellular by hypothesis, and extensions of cellular algebras by cellular ideals are also cellular.) The
cont. We have a Λ n 1 cell net Γ J(Γ) := span{cs,t λ : λ Γ,s,t (λ)}. Now we want to show that Γ Ĵ(Γ) := A n e n J(Γ)A n is a Λ n 1 cell net in A n e n A n. Along the way to doing this we also have to show that J (Γ) := A n An 1 J(Γ) An 1 A n = An e n J(Γ)A n and in particular A n An 1 A n = An e n A n, and if Γ 1 Γ 2, then also J (Γ 1 ) imbeds in J (Γ 2 ). This is a bit tricky because A n e n A n is not a unital algebra and A n is not projective as A n 1 module. and the The
Last slide! and the Now if λ Λ n 1 and Γ 1 Γ 2, with Γ 2 \ Γ 1 = λ, then Ĵ(Γ 2 )/Ĵ(Γ 1 ) = J (Γ 2 )/J (Γ 1 ) = A n An 1 J(Γ 1 )/J(Γ 2 ) An 1 A n = A n An 1 ( λ R ( λ ) ) An 1 A n = (A n An 1 λ ) R ( λ ) An 1 A n ) Now we need that M λ = A n An 1 λ is free as R module, to verify the crucial property in the of a cell net. But as A n module, M λ is Ind( λ ), and by an induction assumption on coherence of the on (A k ) k n, this has a filtration by cell modules for A n, so is free as an R module. The
Last slide + 1 and the The There are also some results about lifting Jucys Murphy elements from Q n to A n in our setting, but I don t want to give details. The method recovers, with a very simple, some results of John Enyang (for Brauer and ) and of Rui and Si in the cyclotomic Brauer and BMW cases.